You are on page 1of 12

WIND ENERGY Wind Energ. (in press) Published online in Wiley Interscience (www.interscience.wiley.com) DOI: 10.1002/we.

231

Research Article

Minimizing Regulation Costs in Multi-area Systems with Uncertain Wind Power Forecasts
Elin Lindgren* and Lennart Sder, KTH, Stockholm, Sweden

Key words: frequency control; multi-area modeling; optimization; regulating market

In a power system where there are many fast and/or large changes in generation or consumption, e.g. in a system with large amounts of wind power, it is more complicated to handle the frequency control efciently. Minimizing regulation costs for the system operator in such a system requires the possibility to simulate the frequency control in the time range from minutes to a few hours. In this paper, it is shown how the frequency control during normal operation can be optimized using a multi-area model, without exceeding frequency limits or transmission capacity. The model has also been expanded to include uncertainties in wind power forecasts, which may lead to an increase of the regulating costs.The optimization model is applied to numerical examples to show the impact of wind power on costs for regulating power and to show the value of better wind speed forecasts. No load forecast errors are taken into account. Copyright 2007 John Wiley & Sons, Ltd. Received 14 November 2006; Revised 2 April 2007; Accepted 7 May 2007

Introduction
Power frequency control is necessary to ensure both quality and security in a power system. To avoid frequency deviations, the generation and the consumption must be equal. The frequency decreases when the consumption exceeds the generation and vice versa, and the consumption varies all the time. Different practices for frequency control are described in Arnott et al.1 In the Nordic power system, for instance, the frequency control is handled in two steps: primary control and secondary control.2 Primary control is purchased on longterm contracts, and when an imbalance occurs, the generation automatically starts to change in a few seconds. How much each generator changes its generation during the primary control is dened by the gain, measured in MW/Hz. Afterward, the frequency is stable, but not at the nominal value. Secondary control is, in the Nordic system, handled by taking in regulating bids in the electricity market. Each bid is normally put not later than 30 min before each hour. It includes a price and a volume, and must be possible to activate within 10 min from acceptance.3 All Nordic bids are collected in one list for all participating countries. Bids are typically accepted according to the cost, but transmission limits (within or between countries) must also be considered, which can have an impact on the desired location of the change in generation. When necessary, secondary control is handled by accepting a bid of regulating power. The change in the generation forces the primary control reserves to go back to their initial state and the frequency is then stable at the nominal level again. In many other systems, the secondary control is instead handled automatically by automatic generation control (AGC). With AGC, the change in power production is divided between all generators available for sec* Correspondence to: E. Lindgren, Electric Power Systems, KTH, Teknikringen 33, SE-100 44, Stockholm, Sweden. E-mail: elin.lindgren@ee.kth.se

Copyright 2007 John Wiley & Sons, Ltd.

E. Lindgren and L. Sder

ondary control instead of choosing one bid at the time. When the secondary control is handled automatically, the frequency control can include tertiary control working as the Nordic secondary control, giving the system operator the possibility to reschedule the production in the system to achieve a more economic operation.

Wind Power Impact


Wind power is one of the most efcient renewable energy sources, but there may be problems installing large amounts of wind power for several reasons.4,5 In this paper, the problems of regulating power and reserves will be discussed. The power generation within an hour in a hydrothermal power system is relatively easy to predict and schedule. Installing large amounts of wind power in that system will change the situation. The generated wind power will uctuate as the wind speed changes, and the variations can then be large within an hour. Also, wind power plants are normally used at the maximum possible level using all the energy in the wind and not keeping reserves. It is more difcult to handle the frequency control when there are large amounts of wind power installed in the system, since both the variations of wind power production and load must be balanced, instead of only balancing the load variations. The added wind power will probably cause more regulation, but in some cases, it is also possible to use less regulating power, if the wind power varies in the same way as the load. The additional amount of regulating power required when large amounts of wind power are installed depends on the variation of wind power, but also on the system and its transmission capacities. In situations where the capacity limits are reached, it is necessary to balance the wind variation in the same area as the wind power plants are located. In Sweden, for instance, this may sometimes be a problem, since most of the wind power plants are located in the south, and most of the regulating power is located in the north. These questions are discussed in Axelsson et al.5 and Magnusson et al.6 where possible wind power variations in Sweden are studied.

Power System Modeling


The transmission system operator (TSO) in a deregulated electricity market needs to be able to simulate the frequency control in order to, for instance, forecast necessary regulation for the next hour or to examine the effects from a single regulation. To create a model of the frequency control, a method for power system modeling is necessary. The existing methods for simulation can roughly be divided either into very detailed dynamic simulations, used for studies of transients in the time range of milliseconds or into economic models where the description of the power system is very simplied and the data is based on hourly averages. Consequently, there is a lack of methods in the time range of minutes and a few hours, which can be useful examining the impact on frequency control from fast variations in the wind power production. One such method is described in Bakken et al.7 using repeated load ow calculations every 5 min to make decisions about activation of regulating bids, but without considering, for instance, transmission limits and overload when deciding which bid to use. This paper shows how the frequency control can be optimized using a multi-area model. Optimizing the frequency control means minimizing the regulation costs, without exceeding frequency limits or transmission capacity. The optimal solution is achieved by making the best possible decisions on bid acceptance: when to accept a bid, which bid and what volume, also considering uncertain load and wind speed forecasts. A pay-asbid approach with different up- and down-regulating prices is applied. Both price and location can affect the decision. The optimization is applied to numerical examples, with and without wind power, to compare regulation costs for different situations and show the impact of wind power on costs for regulating power. The example system consists of a multi-area model, which is assumed to be connected with transmission lines of limited transmission capacity, and the wind power is assumed to be located in several areas. An earlier version of this model was described in Lindgren and Sder,8 showing how the frequency control can be modeled, where a simplied model of wind power uncertainties was included.
Copyright 2007 John Wiley & Sons, Ltd. Wind Energ (in press) DOI: 10.1002/we

Minimizing Regulation Costs in Multi-area Systems

Simulation Method
The simulation consists of several steps (see Figure 1) which will be described in detail below. First of all comes the multi-area modeling, which is done once before the optimization begins. Thereafter comes the optimization, concerning how to accept regulating bids/how to handle the frequency control, which is repeated as the wind power forecasts are updated. This is not done in every time step, but typically every hour. When new wind power forecasts are available, the outcome of the previous time steps is evaluated. The nal result is a successive comparison between production plans and actual production, and the corresponding costs.

Multi-area Model
A multi-area model is used to reduce the amount of necessary input, the numbers of constraints in the optimization problem and the calculation time. The multi-area modeling approach is presented in Lindgren and Sder.9 Some data are more relevant on area basis; e.g. transmission limits are usually set between areas and not for each line between the nodes. On the other hand, too much information can be lost if all calculations are made on an area basis. Thus, it is a challenging task to create an accurate multi-area model from a power system consisting of numerous nodes. An important aspect when transferring the node system to the area system is to ensure that the transmission between areas is as close as possible to the transmission from a load ow of the node system, since keeping the transmission limits are important when decisions on bid activation are made. Several methods to aggregate nodes are described in Machowski et al.,10 for instance, but they are all meant to be used in dynamic simulations, which means they are too detailed to be used here. In Lindgren and Sder,9 different levels of node aggregation were examined, combining nodes and areas, but in this paper all data used in the optimization formulation are on area basis. To distribute the ows in the power system in an appropriate way, a DC load ow11 is used to simplify the calculations. This means that the reactances of the node lines are used to determine a matrix for the node system showing the relation between the net production and the transmission. Thereafter, the matrix can be reduced to reect the area system instead. Another possibility, which also can be combined with the procedure above, is to start with the area system, putting all reactances of the area lines equal, and then modifying them while trying to get a load ow in the area system corresponding to the node system. This is repeated until a given tolerance is reached, as shown in the upper part of the owchart in Figure 1.

Optimization Problem Formulation


A challenging problem for the Nordic TSOs is how to manage the frequency control at the end of an hour, e.g. in the morning, when the load is increasing and the new scheduled generation will start at the change of hour (see Figure 2) (C. Bck, personal communication, 2006). This may cause up-regulation at the end of the hour, with the standard approach to accept the bids in price order, but it can also be necessary to use down-regulation in some areas to avoid exceeding of transmission limits. It is also a challenge to determine when it is optimal to activate the regulating power to minimize the cost, and if a more expensive bid is necessary to avoid overload in the system. In some other systems, a common issue can be large differences in inter-area trading which requires extensive ramping on both sides. This is though not a large question in the Nordic system since the common regulating bid list causes accepted changes of inter-area transmission. The following formulation reects the situation in the Nordic system, where two synchronous systems are connected through HVDC links. These links are here modeled as loads, which only change if a bid available in one synchronous system is accepted for use in the other synchronous system. A bid can only be accepted once, and is then assumed to be active until it is deactivated or until end of hour. When a new hour starts, a new bid list is available. Another assumption in this model (real situation in Sweden) is that the TSO also has the possibility to move the scheduled production up to 15 min, instead of accepting bids close to a change of hour, but this also has a cost.3
Copyright 2007 John Wiley & Sons, Ltd. Wind Energ (in press) DOI: 10.1002/we

E. Lindgren and L. Sder

Figure 1. Flowchart for the simulation

The optimization problem is formulated to capture these conditions, and the calculation will in this paper be done for 3 h starting at 05:30 and ending at 08:30. A list of bids is available for up-regulation or downregulation for each hour. This is formulated as a Linear Programming problem (LP-problem), which is implemented in and solved by the General Algebraic Modeling System (GAMS).12
Copyright 2007 John Wiley & Sons, Ltd. Wind Energ (in press) DOI: 10.1002/we

Minimizing Regulation Costs in Multi-area Systems

50.1

50.0

49.9 05:30 06:00 06:30 07:00 07:30 08:00 08:30

Figure 2. Frequency measured a typical morning

The constraints will keep the frequency within its limits, keep the transmission lower than the capacity of the area lines, control the primary (and secondary) reserves and distribute the ows in the system with a DC load ow. Objective The main objective of the optimization is to minimize the cost of all regulations for the TSO, but the objective function is also used to penalize deviations from the nominal frequency, overload of area connections and bid usage. The terms are weighted to control the impact of each term. It is necessary to have these weights since the values of, for instance, frequency deviations or transmission limits are of very different sizes. The applied weights are commented in the numerical example. The objective function is then min z =
b h B t ,hH T

stop start act cb Vb ( ybt ybt )(Thfinal t tb )

+ w1

nN t ch ,hH T

ch ctch Pnt (Thfinal t ) + w2

nN t T

nt

+ w3 ( fit+ + w4 fit ) + w5 (ul+t + ult )


nN t T lL t T start + w6 ybt n B t T

(1)

Constraints on Generation The change of net generation Pnt in area n at time t is dened as the sum of accepted bids plus change of scheduled production plus changed wind power production minus load change. A bid is assumed to be accepted at act time t but the activation time of the power gives that the power is available at t + tb . The same activation time is used for stopping accepted bids. The net generation is then Pnt =
b n B h B

stop start ch wind Vb ybt t act ybt t act + Pnt + Pnt Dnt
b b

(2)

n N , t T h The scheduled change in generation can be moved up to tch time steps earlier or later (this rule is used in Sweden), but the total volume must here be the same. Decisions concerning this is taken by the TSO when this is more economic than acceptance of regulating bids:
plan Pnt = t + tch s = t tch

ch Pns

plan n N , t T : Pnt > 0

(3)

Copyright 2007 John Wiley & Sons, Ltd.

Wind Energ (in press) DOI: 10.1002/we

E. Lindgren and L. Sder

Here, all scheduled production changes are assumed to be positive, but the changes of the plan may only exceed zero up to tch time steps from the scheduled time, which gives
plan ch 0 Pns Pnt plan n N , t T : Pnt > 0 s { t tch , . . . , t + tch}

(4)

For all other areas and time steps, it is necessary to set the change of schedule to zero:
plan ch Pns = 0 n N , t T : Pnt = 0, s T plan ch Pns = 0 n N , t T : Pnt > 0 s { t tch , . . . , t + tch}

(5) (6)

The change in generation to or from the primary reserve Gnt in area n at time t changes in proportion to the gain dn in area n as Gnt = d n
mN

mN

G d

mt

n N , t T
m

(7)

Constraints on Flow The ow in one direction is dened as being equal to the negative ow in the other direction: Fmnt = Fmnt ( m, n) L , t T (8)

The sum of the changes in generation in area n and the change in transmission to and from area n at time t must be equal to zero, as all losses are neglected:
mL L *

Fnmt = Pnt + Gnt

n N , t T

(9)

To obtain ows corresponding to a DC load ow, the elements mln of the matrix M resulting from the multiarea modeling is used. The change in the ow in line l at time t is given by Flt =
nN

ln

( Pnt + Gnt ) l L, t T

(10)

and the total ow in line l at time t is given by Fltot = Flt + t


s :s t T

Fls

(11)

The transmission between the areas is penalized if exceeding the capacity of the lines. Thus,
min max ul ult Fltot ul + ul+t t

l L , t T

(12)

The transmission between the synchronous systems is only changed if bids are activated in the UCTE system: Flt =

B b n nN t , i = 2

stop start Vb ybt t act ybt t act


b b

l LDC , t T

(13)

Constraints on Frequency Since there is no AGC in the system and the secondary control is handled manually, the frequency varies in the system, normally in the interval 50 0.1 Hz. The frequency at time t in synchronous system i differs from the previous frequency at time t 1 with the total change in net generation at time t divided by the total gain of the system i:
Copyright 2007 John Wiley & Sons, Ltd. Wind Energ (in press) DOI: 10.1002/we

Minimizing Regulation Costs in Multi-area Systems

fit = fit 1 + The frequency must stay within its limits:

nN nN

P d

nt

t T

(14)

nt

f0 fit fit f0 + fit+ 0 fit , fit+ fdev

i I , t T i I , t T

(15) (16)

The time deviation is the time difference between a clock driven by the nominal frequency and a clock driven by the actual frequency. It is dened by an integral and is here calculated as dit = The time deviation must stay within its limits: dmax dit dmax i I , t T (18) tstep f0
s :s t T

( fis f0 )

(17)

Constraints on Bids star The bid acceptance is determined from positive variables between 0 and 1. For example, ybt equals 0.5 if 50% stop of bid b is accepted at time t and ybt equals 0.3 if 30% of bid b is stopped at time t, which gives
start 0 ybt 1 b B , t T

(19) (20)

0y

stop bt

1 b B , t T

A bid can only be activated up to 100% of the available volume (this is the rule in Sweden), and the same is valid for deactivation:
t T

y
t T

start bt stop bt

1 b B 1 b B

(21) (22)

All activated bids must also be stopped, at change of hour or earlier, but not before they have been accepted:
s :s t T

start ybs

stop ybs

b B , t T : t < T b B

(23) (24)

s :s t T

s T

start bs

stop ybs s T

A bid can only be accepted the hour it is available and a bid cannot be accepted closer to the end of an hour than the activation time since it is impossible to activate the bid within the hour. Bids and time steps not available are set to zero:
start ybt = 0 b Bh , t T h : h H start act ybt = 0 b Bh , t T : t > Thfinal tb , h H start act ybt = 0 b Bh , t T : t > Thfinal tb , h H 1

(25) (26) (27)

Corresponding constraints are necessary for stopping bids.


stop ybt = 0 b Bh , t T h , h H stop act ybt = 0 b Bh , t T : Thfinal tb , h H

(28) (29) (30)


Wind Energ (in press) DOI: 10.1002/we

stop bt

act = 0 b Bh , t T : Thfinal tb , h H 1

Copyright 2007 John Wiley & Sons, Ltd.

E. Lindgren and L. Sder

Figure 3. Map with areas shown schematically

Case Study
The model has been applied to the Nordic system, modeled as nine areas connected with 11 lines, as shown in Figure 3. Areas 1, 2 and 3 represent Sweden, areas 4 and 5 represent Norway, area 6 represents Finland, areas 7 and 8 represent Denmark and area 9 represents the northern part of Germany. The optimization is made for 3 h divided into 36 time steps indexed t = 1, . . . , 36, starting at 05:30 and ending at 08:30. The load increases linearly in every time step and the scheduled production increase at change of hour (see Figure 4), but it can be moved up to 15 min. The gain is assumed to be constant in all time steps. The bid list used in the simulations is taken from 23 February 2005 (C. Bck, personal communication) and consists of bids located in all areas of Sweden. Concerning applied weights they have been selected to w1 = w2 = w6 = 1, w3 = 105, w4 = 10 and w5 = 102. Changing these weights less than a factor of 10 will only have a small impact on the costs. For the wind power, persistence forecasts are used which can be updated every hour. At the same time, the actual wind power production for the previous time steps is supposed to be known. The wind power data used are taken from Magnusson et al.,6 where a synthetic wind power database is created from historical weather data from 56 sites all over Sweden to determine the impact of integration of 4000 MW wind power in Sweden, which is further investigated in Axelsson et al.5 In this paper, 4000 MW of wind power is assumed in Sweden
Copyright 2007 John Wiley & Sons, Ltd. Wind Energ (in press) DOI: 10.1002/we

Minimizing Regulation Costs in Multi-area Systems

20 000 MW 18 000 16 000 14 000 05:30 06:00 06:30 07:00 time 07:30 Load Scheduled production 08:00 08:30

Figure 4. Load and scheduled production

but there is no uncertain wind power outside Sweden. The database consists of hourly wind power production data from 10 years, which for the simulations described here has been transferred to 5 min data by linear interpolation. As the bid list was from February, the wind power data of February were selected from the database, giving 283 mornings (10 years including leap days). For each day, two cases were run, one case using the shaded loop illustrated in Figure 1 and one case not using it. There are no load forecast errors.

Case 1: Planning with Persistence Forecast


In the rst case, the optimization is run with a persistence wind power forecast, resulting in a plan when to move scheduled production changes and activate regulating bids. To have some margins for unexpected wind power changes, the plan is done to get a frequency above 49.92 Hz and below 50.1 Hz. Thereafter, this production plan is used while rerunning the simulation using the actual wind power production, determining the impact of the prediction error of the wind power and the actual cost of wind power. In the second step, the limits of the frequency have increased, and the frequency must stay between 49.9 and 50.12 Hz.

Case 2: Planning with Perfect Forecast


The other type of simulation is to use a perfect forecast of the wind power production, determining the optimal plan, and not using the possibility to update the forecast. This case is run for comparison, to determine the value of a perfect forecast of the wind power production. Here, the frequency limits are 49.9 and 50.1 Hz.

Case 3: Without Wind Power


Finally, a case without varying wind power was run. It is quite similar to the case with persistence forecasts, but here no extra margin is kept, giving the frequency limits 49.9 and 50.1 Hz.

Results
The perfect forecasts resulted in a cost reduction by approximately 50% for bids and changes of schedule compared to when the persistence forecasts were used (see Table I). The same result appears in the case without wind power, since keeping the margin in Case 1 at change of hour is costly. The explanation to the lower average cost in Case 2 compared to Case 3 can be a result of the smoothing effect of a continuously changing wind power source compared to the large steps in the planned production. It is realistic that the wind power sometimes is a part of the frequency control, meaning it reduces the power imbalance in the system, but that the wind power also sometimes requires additional regulations. In about 20% of the wind power situations, the margins kept in Case 1 were not enough to be able to stay within the frequency limits in Case 2. For these cases, no costs have been calculated, due to infeasible solutions. The typical morning in these examples has up-regulation before the large increase of production at 07:00, and sometimes is also some necessary down-regulation afterward. This is combined with rescheduling, especially between 06:45 and 07:15. As shown in Table I, the cost for using bids is only about 7% of the total regCopyright 2007 John Wiley & Sons, Ltd. Wind Energ (in press) DOI: 10.1002/we

E. Lindgren and L. Sder Table I. Average costs in SEK Case 1 Cost for bids Cost for re-scheduling Total cost 68,200 891,300 959,500 Case 2 39,200 430,200 469,400 Case 3 32,700 438,000 470,700

ulating cost. This may be correct, but it is possible that the costs for changes of schedule have been modeled inaccurately, since these costs have not been available. As expected, there are larger volumes used from the primary power in Case 1, to compensate the wind power changes after the planning.

Discussion and Conclusions


This paper shows how the frequency control can be optimized using a multi-area model. The optimization model was here also expanded to handle several synchronous systems connected with HVDC lines. The optimization is applied to three numerical cases to show how the regulation costs to compensate for uncertainties in the wind power forecasts can be calculated. As shown in the previous section, the costs for cases with uncertain forecasts are doubled, to keep margins and to plan for regulations which may not be necessary. There is also a higher demand for primary reserves, to be able to handle the nal corrections of the generation. Applied to more critical cases, closer to overload of the transmission lines, and with more uncertain parameters than the wind power production, the model can be used for more complex studies. It is also important for future studies to analyze which margins must be kept to handle situations like those who could not be solved here. To summarize, the value of good forecasts is high and should be investigated more. Concerning numerical results it though has to be noted that there is no load uncertainty and the costs depend on the bid list. However, another aspect of the question comes from those who have the opportunity to leave bids on the regulating market. With increasing amounts of wind power in the system, they hope to make an additional prot on an increased regulated volume with higher prices. From this point of view, it could also be interesting to compare the actual volumes used from the regulating list with the actual rescheduled.

Acknowledgements
The contribution of knowledge from Christer Bck at Svenska Kraftnt and the EU-WILMAR-project (Wind Power Integration in a Liberalized Electricity Market) and the nancial support from the Swedish Energy Agency (STEM) are highly appreciated.

Appendix: Nomenclature
Sets
I = {1, 2} N = {1 . . . m, n . . . m} Ni L = {1 . . . . . . L} L* LDC T = {1 . . . s, t . . . T} synchronous systems areas areas in synchronous system i area lines (m, n), m < n area lines (m, n), m > n HVDC lines time steps
Wind Energ (in press) DOI: 10.1002/we

Copyright 2007 John Wiley & Sons, Ltd.

Minimizing Regulation Costs in Multi-area Systems

H = {1 . . . h . . . H} Th Tch B = {1 . . . b . . . B} Bh Bn

hours time steps in hour h time steps available for changes of scheduled production bids bids available in hour h bids located in area n

Parameters
Dnt plan Pnt wind Pnt F0 Vb cb ctch act tb dn ulmin ulmax Thfinal w1 w6 load change in area n at time t [MW] scheduled change in production in area n at time t [MW] change in difference from forecasted wind power production in area n at time t [MW] total transmission in line l at time t = 0 [MW] available volume of bid b [MW] price per volume for bid b [MW] price per volume to move scheduled production to time t [SEK] activation time for bid b [no of time steps] gain in area n [MW/Hz] min transmission limit between area i and j [MW] max transmission limit between area i and j [MW] the last time step in hour h weights used in the objective function

Constants
f0 fdev dmax tstep tch nominal frequency [Hz] maximal deviation from nominal frequency [Hz] maximal time deviation [s] length of each time step [s] maximal time change for scheduled production [no. of time steps]

Variables
Pnt
ch Pnt Gnt fit ft Fltot t start ybt stop ybt dit ult ul+t mn fit fit+ z

sum of activated bids minus change in load in area n at time t, i.e. the net generation in area n at time t except changes of primary reserves [MW] change of scheduled plan in area n at time t [MW] change in activated primary reserve in area n at time t [MW] frequency in synchronous system i at time t [Hz] change in transmission in line l at time t [MW] total transmission in line l at time t [MW] 1 if bid b is accepted at time t, 0 otherwise 1 if bid b is stopped at time t, 0 otherwise time deviation in synchronous system i at time t [s] overload in line l at time t [MW] overload in line l at time t [MW] coefcient for relation between line l and area n min frequency limit [Hz] max frequency limit [Hz] total cost [SEK]
Wind Energ (in press) DOI: 10.1002/we

Copyright 2007 John Wiley & Sons, Ltd.

E. Lindgren and L. Sder

References
1. Arnott I, Chown G, Lindstrom K, Power M, Bose A, Gjerde O, Morll R, Singh N. Frequency control practices in market environments. Quality and Security of Electric Power Delivery Systems, 2003, CIGRE/IEEE PES International Symposium, Montreal, Canada, 810 October 2003. 2. Ingelsson B, Hermansson H, Messing L. Frequency control under de-regulated conditions. Bulk Power System Dynamics and Control IVRe-structuring, Santorini, Greece, 2428 August 1998. 3. Avtal om balansansvar m.m. med balansreglering (in Swedish). [Online]. Available: http://www.svk.se/upload/ 3325/BAavtal2007.pdf. (Accessed 27 March 2007) 4. Holttinen H. The impact of large scale wind power production on the nordic electricity system. PhD Dissertation, VTT publications 554, Espoo, Finland, 2004. 5. Axelsson U, Murray R, Neimane V. 4000 MW wind power in Sweden. Elforsk rapport 05:19, Energimyndigheten, 2005. 6. Magnusson M, Krieg R, Nord M, Bergstrm H. Effektvariationer av vindkraft. Elforsk rapport 04:34, Energimyndigheten, 2004 (in Swedish). 7. Bakken B, Petterteig A, Haugan E, Walther B. Stepwise power owa new tool to analyse capacity shortage and reserve requirements. 15th Power Systems Computation Conference, Lige, Belgium, August 2005. 8. Lindgren E, Sder L. Wind power impact on costs for regulating power in multi-area markets. Fifth International Workshop on Large-Scale Integration of Wind Power and Transmission Networks for Offshore Wind Farms, Glasgow, Scotland, April 2005. 9. Lindgren E, Sder L. Power system modeling for multi-area regulating market simulation. PowerTech2005, St. Petersburg, Russia, June 2005. 10. Machowski J, Bialek J, Bumby J. Power System Dynamics and Stability. John Wiley & Sons: New York, 1997. 11. Wood A, Wollenberg B. Power Generation Operation and Control (2nd edn). John Wiley & Sons: New York, 1996. 12. GAMS, General Algebraic Modeling System. [Online]. Available: http://www.gams.com. (Accessed 21 April 2006)

Copyright 2007 John Wiley & Sons, Ltd.

Wind Energ (in press) DOI: 10.1002/we

You might also like